Research On DON Detection In Flour Based On Spectroscopy

Flour is one of the main food in people's daily diet,and its quality has always been the focus of everyone's attention.In the evaluation indicators of flour,whether the content of vomit toxin(Deoxynivalenol,DON)meets the standard is of great significance in terms of food safety.DON is widely used in raw materials such as wheat,corn,bran,rice bran,etc.It has a great impact on human health and aquaculture.Therefore,the development of an efficient,low-cost,and practical on-line detection method is of great significance for the safe production and processing of flour.Spectral analysis technology has been widely used in many fields because of its non-destructive,fast and efficient advantages.In this paper,the detection method of DON content in flour will be studied based on spectral technology.The main research contents and related results are as follows:(1)Fourier Transform Infrared Spectrometer(FTIR)technology is used to quantitatively detect the DON content in flour.In the process,polar solvents and centrifugal methods were used to enrich the DON in the flour sample,which increased the DON content in the flour.Collect FTIR spectra of samples before and after enrichment,remove the abnormal samples by Mahalanobis method,Kennard-Stone algorithm reasonably divides the sample set,after proper pretreatment of the spectrum,a variety of methods are used to extract the characteristic band,based on BP neural network and partial least Multiplicative regression(PLSR)established a quantitative prediction model.The results show that the characteristic parameter band is extracted by the trend parameter analysis method,and the modeling effect with BP neural network is the best.Before separation and enrichment,the established models Rp2,RMSEP,RPD were 0.89,0.23mg/kg,7.26,after separation and enrichment,the established models Rp2,RMSEP,RPD were 0.95,0.74mg/kg,10.14,respectively The previous prediction accuracy has been significantly improved.It shows that the enrichment treatment and FTIR technology can be used to detect flour with lower DON content.(2)The limit of the DON content of flour in the national standard is lmg/kg,so before the flour processing process,the detection limit can be appropriately increased to 2mg/kg.According to this,the flour can be divided into three levels:mild(DON<lmg/kg),moderate(lmg/kg<DON<2mg/kg),and severe(DON>2mg/kg)pollution,using near infrared spectroscopy(Near Infrared(NIR)technology classifies and detects DON content in flour.Using the trend parameter method and principal component analysis method to extract the characteristic band,compared with the full spectrum band,based on the quadratic discriminant analysis(QDA)and random forest(RF),a multivariate classification model was constructed.The results show that,when using full-spectrum band modeling,although the overall prediction accuracy can reach 85%,there is an overfitting phenomenon.The trend parameter method greatly reduces the sample dimension,and at the same time,it is better than the DON pollution level.The accuracy of the overall classification of the prediction set reached 90%in models built by other methods,indicating that the feature band extraction algorithm can reduce the data dimension while retaining the effective information of the data.It shows that the NIR technology can effectively and quickly classify the DON pollution level of flour.(3)Select the ocean optical spectrometer,and use the characteristic wave band extracted by the trend parameter method in the previous step and the established QDA classification and discrimination model.Based on this complete TP-QDA classification model and Matlab-GUI platform,a flour vomiting toxin near-infrared spectrometer was developed,which can predict the DON level in flour based on the near-infrared spectroscopy data of flour,and can quickly classify flour in a certain range.After the instrument is completed,a new batch of samples is selected to check its accuracy.The results show that the prediction accuracy of moderate pollution levels can reach 93.8%,the overall prediction accuracy is 82.14%,and the instrument is easy to carry,the software is easy to deploy,and the overall use is convenient.Have the conditions for promotion and use.The above results show that the detection of flour DON content based on spectroscopic technology is a fast and effective method.The software developed by the spectrometer combined with the Matlab-GUI platform has the advantages of portability and low price,which is conducive to improving the level of food safety and quality in China.Work efficiency has corresponding reference value.